My City, My Voice: Listening to the Citizen Views from Web Sources

Abstract

To facilitate an environment of inclusive urban management, civic agencies need to listen to the voices of citizens on web sources such as social media, online blogs, public forums and so on. Owing to the vastness and noisy nature of online data, it is challenging, yet important to mine actionable issues related to a city as faced by the citizens firsthand, so that timely measures can be taken by the administration to remedy them. In this work, we filter, analyze, and model web data on urban civic issues of a city, with respect to three modalities - semantics, spatial and temporal. We have come up with a novel approach that captures the contexts through dense distributed word embedding as well as identifies the latent issues through a generative model. Due to the scarcity of geo-tagged posts and delayed reporting, we rely primarily on the textual content of the data for location mining and temporal resolution. We present a first-of-a-kind unified system named CUrb that introduces a novel pipeline to construct long term topology of issues across three dimensions, aggregated over a variety of documents. Through extensive experimentation, we demonstrate the efficacy of our system both qualitatively and quantitatively. It achieves improvement upto 24% compared to the state-of-the-art technique.